mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Hideo AOKI <aoki@sdl.hitachi.co.jp>
To: Rik van Riel <riel@redhat.com>
Cc: Andrew Morton <akpm@osdl.org>, linux-kernel@vger.kernel.org
Subject: Re: 2.6.8-rc2-mm2
Date: Thu, 05 Aug 2004 19:56:37 +0900	[thread overview]
Message-ID: <41121265.8000909@sdl.hitachi.co.jp> (raw)
In-Reply-To: <Pine.LNX.4.58.0408020806210.13053@dhcp030.home.surriel.com>

Hi, 

Rik van Riel wrote:

> I would really appreciate any testing results on this, both good
> and bad.  I want to get this thing tuned and into a generally good
> shape for use by everybody upstream.
> 
> I'm especially interested in how this affects compute servers,
> desktops and heavily overloaded network servers (the "spamassassin
> slowed my system to a crawl" symptom would be one to test ;)).
> 
> I suspect the patch may need some tweaking to help interactivity
> in some cases, but maybe it'll already work magically by itself...

To evaluate your patch, I tested -mm kernels for thrashing generated
by our benchmark suit.

In 2.6.8-rc2-mm2, performance of application increased, and overhead
of kernel decreased.


** Overview of workload

 Benchmark program creates disk I/O processes and shared memory access
 processes. Generating disk I/O and shared memory access at the same
 time.



** Detailed environment of performance evaluation

 - Hardware
   CPU: Xeon 1.6GHz * 4
   Memory: 2GB
   HDD: IDE ATA100 

 - The benchmark suit 
   The benchmark suit is the wblg-disk, which I have released on
   SourceForge.

   <http://sourceforge.net/project/showfiles.php?group_id=110454&package_id=119281> 
   # I used alpha version (1.0.3-alpha) in this test.
   # I will release 1.0.3 soon.

 - configuration of benchmark
   1. disk IO workload
     + number of process: 4 processes
     + read/write ratio: read 0% / write 100%
     + file IO size: 1KB - 256KB (random) 

   2. memory access workload
     + number of process: 32 processes
     + Shard memory regions: 512MB * 4 
     + Each process accesses one shared memory region.
       (8 processes share one shared memory region.)
     + Each process repeats 4 byte memory access each 4KB.

   Measurement time: 1 hour
   Measurement items: write throughput

 - Other configurations
   Using Oprofile 
   Issue vmstat command each 1 minute



** Results of performance evaluation

* result of benchmark
                        write throughput [MB/s]
        2.6.8-rc2-mm1:  2.39
        2.6.8-rc2-mm2:  2.68


* result of vmstat
                        swap in(kB/s)   swap out(kB/s)
        2.6.8-rc2-mm1:  0 - 641         0 - 583         
        2.6.8-rc2-mm2:  0 - 727         0 - 684         


* results of Oprofile

  - System profiling (top 2)

    Overhead of kernel decreased by 3.9%.

    2.6.8-rc2-mm1:
        No.     samples         %               
        1.      170981153       77.6102 wblg-disk_client
        2.       46808551       21.2469 vmlinux-2.6.8-rc2-mm1

    2.6.8-rc2-mm2:
        No.     samples         %               
        1.      179798682       81.6102 wblg-disk_client
        2.       38141112       17.3122 vmlinux-2.6.8-rc2-mm2


  - Kernel profiling (top 10) 

    Ratio of page_referenced_one() and kmap_atomic() decreased.

    2.6.8-rc2-mm1:
        No.     samples         %       symbol name
        1       11398599        24.3515 page_referenced_one
        2        5652276        12.0753 kmap_atomic
        3        2257662         4.8232 sysenter_past_esp
        4        2098275         4.4827 flush_tlb_others
        5        1467288         3.1347 __copy_to_user_ll
        6        1167740         2.4947 page_address
        7        1145108         2.4464 page_fault
        8         946909         2.0229 flush_tlb_page
        9         889151         1.8995 do_gettimeofday
        10        828174         1.7693 refill_inactive_zone

    2.6.8-rc2-mm2:
        No.     samples         %       symbol name
        1        6057840        15.8827 page_referenced_one
        2        3707571         9.7207 kmap_atomic
        3        2531840         6.6381 sysenter_past_esp
        4        1707017         4.4755 __copy_to_user_ll
        5        1328610         3.4834 flush_tlb_others
        6        1233726         3.2346 page_fault
        7        1140917         2.9913 do_gettimeofday
        8         970534         2.5446 get_offset_tsc
        9         759579         1.9915 mark_offset_tsc
        10        650208         1.7047 page_address


Best regards,
Hideo AOKI

Systems Development Laboratory, Hitachi, Ltd.


  reply	other threads:[~2004-08-05 10:59 UTC|newest]

Thread overview: 55+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2004-08-02  8:55 2.6.8-rc2-mm2 Andrew Morton
2004-08-02  9:23 ` 2.6.8-rc2-mm2 Espen Fjellvær Olsen
2004-08-02 12:10 ` 2.6.8-rc2-mm2 Rik van Riel
2004-08-05 10:56   ` Hideo AOKI [this message]
2004-08-05 12:55     ` 2.6.8-rc2-mm2 Hideo AOKI
2004-08-02 13:07 ` 2.6.8-rc2-mm2 Con Kolivas
2004-08-03  6:43   ` 2.6.8-rc2-mm2 Nick Piggin
2004-08-03  7:35     ` 2.6.8-rc2-mm2 Con Kolivas
2004-08-03  7:43       ` 2.6.8-rc2-mm2 Nick Piggin
2004-08-03  7:45       ` 2.6.8-rc2-mm2 Con Kolivas
2004-08-04  8:35         ` 2.6.8-rc2-mm2 Nick Piggin
2004-08-04 10:12           ` 2.6.8-rc2-mm2 Con Kolivas
2004-08-04 10:33             ` 2.6.8-rc2-mm2 Nick Piggin
2004-08-04 11:42               ` 2.6.8-rc2-mm2 bert hubert
2004-08-05  1:21                 ` 2.6.8-rc2-mm2 Nick Piggin
2004-08-04 10:31           ` 2.6.8-rc2-mm2 Ingo Molnar
2004-08-04 10:46             ` 2.6.8-rc2-mm2 Nick Piggin
2004-08-04 10:47             ` 2.6.8-rc2-mm2 Con Kolivas
2004-08-04 12:45               ` 2.6.8-rc2-mm2 Ingo Molnar
2004-08-05  2:53                 ` 2.6.8-rc2-mm2 Con Kolivas
2004-08-05  8:34                   ` 2.6.8-rc2-mm2 Helge Hafting
2004-08-05  9:34                     ` 2.6.8-rc2-mm2 Con Kolivas
2004-08-05 10:47                   ` 2.6.8-rc2-mm2 Ingo Molnar
2004-08-02 13:52 ` 2.6.8-rc2-mm2 William Lee Irwin III
2004-08-02 16:12   ` 2.6.8-rc2-mm2 William Lee Irwin III
2004-08-02 14:28 ` 2.6.8-rc2-mm2 William Lee Irwin III
2004-08-02 15:31 ` 2.6.8-rc2-mm2 Mark Watts
2004-08-02 15:33   ` 2.6.8-rc2-mm2 Jesper Juhl
2004-08-02 15:38   ` 2.6.8-rc2-mm2 Dominik Karall
2004-08-02 16:02     ` 2.6.8-rc2-mm2 Mark Watts
2004-08-03  6:55   ` 2.6.8-rc2-mm2 Clemens Schwaighofer
2004-08-02 15:57 ` 2.6.8-rc2-mm2 Martin Zwickel
2004-08-02 17:59 ` 2.6.8-rc2-mm2 (compile stats) John Cherry
2004-08-02 18:15 ` 2.6.8-rc2-mm2 Peter Osterlund
2004-08-02 22:16 ` [patch] 2.6.8-rc2-mm2: compile error with SWAP=n Adrian Bunk
2004-08-03  8:27 ` 2.6.8-rc2-mm2 Mark Watts
2004-08-03 13:05   ` 2.6.8-rc2-mm2 R. J. Wysocki
2004-08-03 13:13 ` 2.6.8-rc2-mm2 Con Kolivas
2004-08-03 19:44 ` 2.6.8-rc2-mm2 Adrian Bunk
2004-08-04 16:32 ` 2.6.8-rc2-mm2 Dominik Karall
2004-08-05  4:38 ` [sparc32] [1/13] turbosparc flush warnings William Lee Irwin III
2004-08-05  4:39   ` [sparc32] [2/13] sparc32 init_idle() William Lee Irwin III
2004-08-05  4:41     ` [sparc32] [3/13] sun4d cpu_present_map is a cpumask_t William Lee Irwin III
2004-08-05  4:44       ` [sparc32] [4/13] smp_processor_id() BITFIXUP fixes William Lee Irwin III
2004-08-05  4:46         ` [sparc32] [5/13] reinstate smp_reschedule_irq() William Lee Irwin III
2004-08-05  4:47           ` [sparc32] [6/13] remove references to start_secondary() William Lee Irwin III
2004-08-05  4:48             ` [sparc32] [7/13] remove references to num_context_patch2 William Lee Irwin III
2004-08-05  4:49               ` [sparc32] [8/13] define cache_decay_ticks William Lee Irwin III
2004-08-05  4:54                 ` [sparc32] [9/13] remove unused variable in dvma.c William Lee Irwin III
2004-08-05  4:55                   ` [sparc32] [10/13] sun4 does not support SMP William Lee Irwin III
2004-08-05  4:56                     ` [sparc32] [11/13] make CONFIG_SMP depend on CONFIG_BROKEN William Lee Irwin III
2004-08-05  5:01                       ` [sparc32] [12/13] gcc-3.3 macro parenthesization fix for memcpy.S William Lee Irwin III
2004-08-05  5:31                         ` [sparc32] [13/13] ignore undefined symbols with 3 or more leading underscores William Lee Irwin III
2004-08-05  7:28                         ` [sparc32] [12/13] gcc-3.3 macro parenthesization fix for memcpy.S Jakub Jelinek
2004-08-05  7:38                           ` William Lee Irwin III

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=41121265.8000909@sdl.hitachi.co.jp \
    --to=aoki@sdl.hitachi.co.jp \
    --cc=akpm@osdl.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=riel@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®